Capterra Scraper

Scrape Capterra software reviews from public product pages and export structured review data to JSON, CSV, or Excel. This Capterra review scraper follows paginated regional product pages to collect ratings, pros, cons, reviewer details, and direct review links, providing a practical Capterra data export for research and reporting.

Enter a product URL containing a numeric product ID, set a review limit, and run the actor. A validated input using a US Capterra URL returned 40 distinct reviews on September 26, 2026; the golden check required at least 30 and passed. The actor converts the product ID to the equivalent Canadian product route because the tested US product and review pages returned HTTP 403. The Canadian page exposed the same product and its public Capterra review cards, with pagination beyond the first page.

What data can you extract from Capterra?

Each dataset item represents one review card. The output is flat so it can be read in spreadsheets and joined to other datasets by reviewId. Text fields can be null when a reviewer did not supply them. Scores remain numbers, and writtenOn is the date shown on the public regional page in YYYY-MM-DD form.

Field	Description	Example
productUrl	Product URL supplied in the input	https://www.capterra.com/p/141633/PowerPoint/reviews/
sourceUrl	Regional product page used to retrieve the cards	https://www.capterra.ca/software/141633/powerpoint
productName	Name displayed on that product page	Product display name
reviewId	Capterra review identifier	Capterra___7216923
globalReviewId	Shared identifier used for deduplication	Capterra___7216923
reviewUrl	Public link anchored to the individual review	Product URL with review= parameter
title	Review headline	The Industry Standard for Professional Presentations
writtenOn	Review date displayed on the regional page	2026-09-22
overallRating	Overall score on a five-point scale	5
easeOfUseRating	Ease of use sub-rating	5
customerSupportRating	Service or support sub-rating	5
functionalityRating	Features sub-rating	5
valueForMoneyRating	Value sub-rating	5
recommendationRating	Likelihood to recommend, from 0 to 10	9
prosText	Positive review text	Reviewer supplied text
consText	Negative review text	Reviewer supplied text
generalComments	Overall experience paragraph	Text or null
adviceToOthers	Advice field if available	null on the validated page
sourceSite	Review origin label	Capterra
vendorResponseText	Public vendor response, if present	Text or null
reviewerName	Public display name or anonymous label	Reviewer display label
reviewerJobTitle	Public occupation or role	Virtual Assistant
reviewerCountry	Country shown with the reviewer role	Brazil
reviewerCompanySize	Public employee range or size label	Self Employed
reviewerIndustry	Public industry label	Marketing & Advertising
reviewerTimeUsedProduct	Reported duration of use	2+ years
reviewerProfilePicUrl	Public profile image link	URL or null
reviewerVerified	Whether a visible verification marker is shown	true
pageNumber	Page on which the review card appeared	1
scrapedAt	UTC time when the record was exported	ISO timestamp

The four sub-ratings measure different aspects of the product. Do not substitute one for another when a value is absent. The recommendation field uses a ten-point scale and is distinct from the five-point overall score. A missing vendor response means that the response was not present in that extracted card; it does not prove that no response exists elsewhere. reviewerVerified repeats a visible source marker and is not an independent identity check.

How to scrape Capterra reviews

1. Copy a Capterra product URL that contains its numeric product ID. A review URL and a product overview URL both work when they contain the ID and slug.
2. Paste the URL into productUrl. Set maxResults to the number of reviews you want, or use 0 to continue through all publicly reachable pages. Begin with a small number when checking an unfamiliar product.
3. Choose a sort order and, if needed, one or more star ratings. The source provides most recent and most helpful ordering. Rating-order choices are applied within each retrieved page.
4. Start the actor. Inspect the resulting dataset count and fields, then download JSON, CSV, or Excel from the run page.

The actor starts with page 1 and follows the numeric pagination links shown on the product page. In the validated product, the first page showed five featured reviews and later pages showed 25 reviews each. The first 40 results therefore came from three pages. The actor deduplicates records by reviewId as it moves through pages and stops when the requested count, last page, or run budget is reached.

Input

Parameter	Type	Default	Description
productUrl	string	Required	Capterra product or review URL with numeric product ID and slug.
maxResults	integer	100	Maximum reviews to write. 0 removes the actor-side item cap.
sort	string	MOST_RECENT	Most recent, most helpful, highest rated per page, or lowest rated per page.
starRating	array of strings	Empty	Only include selected overall ratings from 1 to 5.
proxyConfiguration	object	Platform proxy	Override public-page proxy routing when necessary.

A realistic input is:

{
  "productUrl": "https://www.capterra.com/p/141633/PowerPoint/reviews/",
  "maxResults": 40,
  "sort": "MOST_RECENT"
}

That exact input produced 40 items in the September 26, 2026 golden run. The source can change its page layout, availability, or review count over time, so a later run can differ. If a page is blocked or does not contain review cards, the run reports an error rather than silently treating the product as having no reviews.

The star filter checks each card's overall rating. The actor also sends the lowest selected rating as a source threshold to narrow the pages when possible. It still checks exact selected values before export, because a threshold of four includes both four-star and five-star reviews. When selecting rare ratings, the actor may need to read many pages before it finds enough matches. A requested count is an upper bound, not a promise that the source contains that many matching reviews.

Output example

This is a trimmed subset of a real item from the successful golden run. Public reviewer text, name, image URL, and the run timestamp are omitted here for brevity; they are present in the dataset when the source supplies them.

{
  "productUrl": "https://www.capterra.com/p/141633/PowerPoint/reviews/",
  "sourceUrl": "https://www.capterra.ca/software/141633/powerpoint",
  "reviewId": "Capterra___7216923",
  "globalReviewId": "Capterra___7216923",
  "reviewUrl": "https://www.capterra.ca/software/141633/powerpoint?review=Capterra___7216923#highlighted-review",
  "title": "The Industry Standard for Professional Presentations",
  "writtenOn": "2026-09-22",
  "overallRating": 5,
  "easeOfUseRating": 5,
  "customerSupportRating": 5,
  "functionalityRating": 5,
  "valueForMoneyRating": 5,
  "recommendationRating": 9,
  "sourceSite": "Capterra",
  "reviewerCountry": "Brazil",
  "reviewerCompanySize": "Self Employed",
  "reviewerTimeUsedProduct": "2+ years",
  "reviewerVerified": true,
  "pageNumber": 1
}

The validation check required reviewId, reviewUrl, overallRating, title, writtenOn, sourceSite, and reviewerName on every item; all were filled in the 40-item run. prosText and consText were also filled on every item, above the required 90 percent threshold. These results establish that the actor handled the tested product and pages at that time. They do not prove that every product exposes the same optional fields or that all historical pages will always be reachable.

Use cases

• Product feedback analysis: Group review text into recurring benefits and problems while keeping the numeric scores alongside each comment.
• Competitor research: Compare public reviews across several software products using the same stable field names, with one product URL per run.
• Quality reporting: Monitor recent review dates and scores, then inspect the full pros and cons before drawing a conclusion from a rating trend.
• Support assessment: Identify reviews that discuss service and compare the support sub-rating with the reviewer's description when both are present.
• Dataset enrichment: Join reviews to an existing product catalog using the input product URL, source URL, and unique review ID.
• Ongoing collection: Schedule periodic runs and deduplicate by reviewId to retain newly visible cards without storing repeat copies.

For trend work, keep writtenOn and scrapedAt separate. The former is the date displayed for the review; the latter is when the actor exported it. A review can become visible to the actor after its publication date, and rerunning a page later does not change the publication date. For cross-product comparisons, use the same sort and star choices for each product and check how many reviews were actually collected.

How much does it cost to scrape Capterra?

The configured result price is $0.30 per 1,000 exported reviews, or $0.0003 per result. The price is charged on items written to the dataset, so setting a large maxResults value alone does not create result charges. The displayed price in the actor input is the authoritative current price. The platform may separately account for compute, transfer, and proxy usage in the run's cost details.

At the configured result price, 40 reviews cost $0.012 in result events, 1,000 reviews cost $0.30, and 5,000 reviews cost $1.50. The successful 40-item golden run reported about $0.0021 in platform usage; that is a different figure from the event-price calculation. Costs depend on how many pages must be read, especially when a star filter selects rare reviews. The arithmetic examples assume those quantities are available and collected; the public source decides the actual reachable count.

A quick check with maxResults set to 30 reads only enough pages to reach that output cap. A full-history request with maxResults set to zero can read many pages, so decide whether that scope is necessary. The validated product's page linked as far as page 837. This is a product-specific observation, not a universal cap. Each additional page requires a request, and long runs are more exposed to changes in site availability or layout.

Tips for best results

Supply the canonical product URL when possible. The actor extracts the numeric ID and slug and builds the corresponding regional route, so extra tracking parameters and a trailing /reviews/ segment do not affect which product it targets. Confirm that the product name shown in the dataset matches your intended product, particularly when a URL has been copied from an older listing.

Use MOST_RECENT to collect reviews in the site's recent order. HIGHEST_COMPLETENESS_SCORE follows the site's default most-helpful order. HIGHEST_RATED and LOWEST_RATED reorder cards within each retrieved page; they do not globally sort every review across all pages. If you need a strict global ranking, collect the desired range and sort the exported dataset afterward.

For a star-specific analysis, select only the ratings needed and inspect the output count. The source's rating filter is a lower-bound threshold, so the actor performs an exact check before emitting rows. A low-star request can require much deeper pagination than a general request. If the run ends with fewer reviews than requested, inspect the source's available count and the run log before concluding that the product has no more matching reviews.

Use reviewId to deduplicate across repeated runs. This identifier remains the best key for identifying the same public review when it appears on different pages or on subsequent collection dates. The actor deduplicates within a run, but downstream storage should also enforce uniqueness when runs are combined. Preserve the direct reviewUrl for audit and the original productUrl for grouping.

Limits

The tested US product and review paths returned HTTP 403 through the available network routes. The actor therefore reads public Canadian product pages with matching numeric product IDs. Those pages returned the review cards used in the successful golden check. Availability can vary by product or time. If the regional page fails, the actor cannot guarantee a second public source with equivalent detail.

The regional pages place review data in HTML. Their JSON-LD contains a shorter review representation but omits the review ID, title, separate pros and cons, sub-ratings, and reviewer profile needed for this output. The actor uses the visible review cards for those fields. A site redesign can change the card structure or pagination and require an update. The last page is determined from links on the current product page; the observed page count is not assumed for other products.

The actor does not infer absent switching reasons, alternative products, incentive details, or full vendor-response objects. adviceToOthers and vendorResponseText can be null when those fields are not present in a card. The visible date is a local display date and can differ by a day from a separate UTC timestamp exposed in page metadata. Dates are therefore exported as displayed calendar dates, without an invented time of day.

Most recent and most helpful sorting follow source URLs. Rating-based sorting applies per page. A multi-star filter narrows the requests with a lower-bound rating and then applies exact matching locally. These details matter if you use a small maxResults value: the selected subset can reflect page order and filtering, not a global ranking of every review ever submitted.

Automated workflows

A scheduled workflow can pass a product URL and maximum count, read the default dataset, and upsert rows keyed by reviewId. It should inspect the run status and item count before refreshing a report. A second workflow can run several products independently, add product catalog identifiers downstream, and keep the extraction time to record when each review was observed.

For monitoring, compare the newest writtenOn value with the prior run and flag new IDs. For quality checks, verify that overallRating, prosText, and consText meet your required fill rates before calculations are published. These checks are especially useful after a source layout change: a successful HTTP response alone does not prove that all review fields were extracted correctly.

FAQ

Can it collect more than the first page?

Yes. The validated product had five cards on page 1 and 25 on each tested later page. The actor followed pagination and exported 40 distinct reviews from three pages. It stops at the requested count or the final page, and it respects the run budget when writing results.

Why is the source URL on a different regional domain?

The tested US product pages returned HTTP 403. The corresponding public Canadian product route was reachable and contained the review cards. The dataset retains the URL you entered in productUrl and records the page it read in sourceUrl.

Are highest and lowest rated sorts global?

No. Those choices reorder the cards on each retrieved page. Collect a sufficiently broad dataset and sort the export if you need a global ranking. Most recent and most helpful use the site's available page ordering.

What happens when a field is empty?

The public card did not supply that field or the field was not visible in the checked layout. Missing sub-ratings should remain missing rather than be treated as zero. An absent vendor response does not establish that no response exists on another page.

Do I need a proxy?

The actor uses platform proxy routing by default and escalates when a request is blocked. The regional page was accessible in the validation run. Reachability can vary by route and time, and a proxy cannot guarantee that a site will continue serving a page.

Can I export to a spreadsheet or schedule runs?

Yes. The default dataset can be downloaded as CSV or Excel, and runs can be scheduled. Deduplicate records by reviewId when appending results from multiple runs to the same spreadsheet or data store.

Is it legal to use these reviews?

Use public review data for legitimate purposes, respect site terms and applicable privacy rules, and avoid misuse of reviewer details. The dataset can include public names and profile links; retain only the fields needed for your use case.

Integrations

The default dataset works with the platform API, scheduled runs, webhooks, CSV or Excel exports, and common data pipelines. A reliable integration checks run success, confirms the row count, and stores reviewId, productUrl, and scrapedAt with each record. For multiple products, run each URL separately and combine datasets after checking that product names and IDs match your catalog.

Support

Open an issue on the Issues tab with the public product URL, run ID, and expected versus actual item count. Include a review link when reporting a missing field so the visible card can be compared with the exported record. Field requests can be reviewed when the public source provides the data.